gear (version 0.1.5)

mle: Finds maximum likelihood estimates of model parameters for a geostatistical model

Description

mle estimates the parameters of a geostatistical linear model. The function is written to automatically adapt based on the class of object. See Details.

Usage

mle(object, reml = FALSE, est = "e", ...)
# S3 method for geolmStd
mle(
  object,
  reml = FALSE,
  est = "e",
  lower = NULL,
  upper = NULL,
  method = "nlminb",
  itnmax = NULL,
  control = list(),
  ...
)

Arguments

object

A geostatistical linear model object producted by the geolm function.

reml

A logical value indicating whether standard maximum likelihood estimation should be performed (reml = FALSE). If reml = TRUE, then restricted maximum likelihood is performed. Defaul is FALSE.

est

A character vector indicator whether the error variance (est="e") or finescale variance (est = "f") should be estimated. The other component of the nugget variance is held constant, and in the case of a geolmStd object, is set to 0.

...

Currently unimplemented.

lower

A vector of 2 or 3 specifying the lowerbound of parameter values. See Details.

upper

lower A vector of 2 or 3 specifying the lowerbound of parameter values. See Details.

method

The optimization method. Default is "nlminb", with "L-BFGS-B" being another acceptable choice. See optimx for details.

itnmax

An integer indicating the maximum number of iterations to allow for the optimization prodedure.

control

A list of control parameters passed internally to optimx. See optimx for details.

Details

In the case of a geolmStd object, the likelihood has been concentrated so that only the range parameter r and a scale parameter lambda = nugget/psill need to be estimated.

If object is a geolmStd, then lower is of length 2 if the covariance model of cmod is not matern or amatern. Otherwise, it should be of length 3. The first parameter is related to the range parameter r, the second to the scale parameter lambda, and the third to par3, if applicable. If lower = NULL, then the lower bounds are 0.001, 0, and 0.1, respectively. A similar pattern holds for upper, with the default being 3 * max(d), where d is the matrix of distances between coordinates, 5, and 2.5.

The kkt argument in the control list is set to be FALSE.

Examples

# NOT RUN {
set.seed(10)
n = 100
# }